[patch, commited] for PR 32949

[Zdenek Dvorak]

Hello,

for variables in nowrap types whose evolution is [something, + step]
and step is not constant, we currently claim that they may wrap.  This
patch moves the cheking that the step is a constant (that is necessary
for analysis of wrapping based on the estimates on number of iterations
of loop) after the type check (that is correct even if the step is a
loop-invariant variable).

Bootstrapped & regtested on i686, commited.

Zdenek

	PR tree-optimization/32949
	* tree-ssa-loop-niter.c (scev_probably_wraps_p): Test nowrap_type_p
	before failing for ivs with non-constant step.

Index: tree-ssa-loop-niter.c
===================================================================
*** tree-ssa-loop-niter.c	(revision 127674)
--- tree-ssa-loop-niter.c	(working copy)
*************** scev_probably_wraps_p (tree base, tree s
*** 2969,2976 ****
       2032, 2040, 0, 8, ..., but the code is still legal.  */
  
    if (chrec_contains_undetermined (base)
!       || chrec_contains_undetermined (step)
!       || TREE_CODE (step) != INTEGER_CST)
      return true;
  
    if (integer_zerop (step))
--- 2969,2975 ----
       2032, 2040, 0, 8, ..., but the code is still legal.  */
  
    if (chrec_contains_undetermined (base)
!       || chrec_contains_undetermined (step))
      return true;
  
    if (integer_zerop (step))
*************** scev_probably_wraps_p (tree base, tree s
*** 2981,2986 ****
--- 2980,2990 ----
    if (use_overflow_semantics && nowrap_type_p (type))
      return false;
  
+   /* To be able to use estimates on number of iterations of the loop,
+      we must have an upper bound on the absolute value of the step.  */
+   if (TREE_CODE (step) != INTEGER_CST)
+     return true;
+ 
    /* Don't issue signed overflow warnings.  */
    fold_defer_overflow_warnings ();